An Explainable EEG Framework for Alzheimer’s and Frontotemporal Dementia Diagnosis Using Pairwise CSP and Random Forest

Alzheimer’s and frontotemporal dementia are generally recognized as common brain diseases. These serious diseases frequently change the brain’s connections and functions. They significantly impact well over 100 million people worldwide, and the number is growing. This study introduces an innovative technique for analyzing complex brain signals. This unique method is officially named Pairwise Common Spatial Pattern or PCSP. PCSP significantly enhances the accuracy of classification for EEG signals. Six different classifiers were rigorously evaluated. Random Forest demonstrated the highest overall precision, achieving an accuracy of 97.75%. Explainable AI (XAI) techniques were employed, with SHAP and LIME effectively identifying critical features and relevant EEG channels. These metrics are highly beneficial for this classification. The entire system definitively demonstrated rapid processing, completing in about 1.07 ms per sample. This speed indicates its potential use in accurate real-time applications. This novel approach establishes an efficient method for accurately diagnosing healthy individuals. It also facilitates earlier diagnosis for patients with devastating Alzheimer’s and frontotemporal dementia.